Basic and Applied Ecology 27 (2018) 83–95
Edge contrast does not modulate edge effect on plants and pollinators E. Andrieua,∗ , A. Cabanettesa , A. Alignierb , I. Van Halderc , D. Alardd , F. Archauxe , L. Barbaroa,d , C. Bougete , S. Baileye , E. Corcketd , M. Deconchata , M. Vigana , A. Villemeye , A. Ouina a
DYNAFOR, INRA, INP Toulouse, 24 Chemin de Borde Rouge, 31326 Castanet Tolosan, France BAGAAP, INRA, Agrocampus Ouest, 65, rue de Saint-Brieuc, 35042 Rennes, France c BIOGECO, INRA, Univ. Bordeaux, 33610 Cestas, France d BIOGECO, INRA, Univ. Bordeaux, 33615 Pessac, France e Irstea, UR EFNO, Domaine des Barres, 45290 Nogent-sur-Vernisson, France b
Received 21 February 2017; received in revised form 22 November 2017; accepted 23 November 2017 Available online 29 November 2017
Abstract Edge contrast, is one of the main determinants of edge effects. This study examines the response of plant and pollinator diversity (bees and butterflies) to forest edge contrast, i.e. the difference between forests and adjacent open habitats with different disturbance regimes. We also investigated a potential cascading effect from plants to pollinators and whether edge structure and landscape composition mediate the relationship between edge contrast and beta diversity of pollinators. We sampled 51 low-contrast edges where forests were adjacent to habitats showing low levels of disturbance (i.e. grey dunes, mowed fire-breaks, orchards, grasslands) and 29 high-contrast edges where forests were adjacent to more intensively disturbed habitats (i.e. tilled firebreaks, oilseed rape) in three regions of France. We showed that plant diversities were higher in edges than in adjacent open habitat, whatever the edge contrast. However, plant beta diversity did not differ significantly between low and high-contrast edges. While we observed higher pollinator diversities in adjacent habitats than in low-contrast edges, there were no significant differences in pollinator beta diversity depending on edge contrast. We did not observe a cascading effect from plants to pollinators. Plant and bee beta diversities were mainly explained by local factors (edge structure and flower cover) while butterfly beta diversity was explained by surrounding landscape characteristics (proportion of land cover in grassland). ¨ © 2017 Gesellschaft f¨ur Okologie. Published by Elsevier GmbH. All rights reserved.
Keywords: Forest edges; Butterflies; Wild bees; Beta diversity; Landscape composition; France
Introduction Most of Europe’s forest cover is made up of fragmented forests (Larsson, 2001) resulting mainly from land-use ∗ Corresponding
author. E-mail address:
[email protected] (E. Andrieu).
changes, like habitat conversion for agriculture but also reforestation of agricultural land over historical time (Tilman et al., 2001; Dupouey, Dambrine, Laffite, & Moares, 2002; Andrieu, Ladet, Heintz, & Deconchat, 2011), and road network (Ibisch et al., 2016). These small forests are characterized by their small area, their isolation from other forests and their high edge/core-area ratio. Forest edges are defined
https://doi.org/10.1016/j.baae.2017.11.003 ¨ 1439-1791/© 2017 Gesellschaft f¨ur Okologie. Published by Elsevier GmbH. All rights reserved.
84
E. Andrieu et al. / Basic and Applied Ecology 27 (2018) 83–95
as the transition zone between open habitats and forests (Matlack & Latvaitis, 1999). They can present a great variability in their three-dimensional structure such as their width, their shape or tree stem density (Essen, Ringvall, Harper, Christensen, & Svensson, 2016), and the quantity and quality of available habitats in edges depend in part on their structure (Didham & Lawton, 1999; Ries, Fletcher, Battin, & Sisk, 2004). Forest edge characteristics are also under human control: they are mainly managed by woodlot owners and farmers to prevent tree growth from encroaching on the field or shading the crop and to specific cutting operations for firewood (Du Bus de Warnaffe, Deconchat, Ladet, & Balent, 2006), and they are also subject to disturbance from forestry and adjacent farming practices such as tillage, pesticide drift or mowing (Kleijn & Verbeek, 2000). Several local factors can influence the magnitude of edge effects: edge orientation and structure, the quality and quantity of resources and/or refugia found in edges and their adjacent habitats and edge contrast (Ries et al., 2004). Edge contrast expresses the differences in quality and/or vegetation structure (height, density) between the two adjacent habitats forming the edge (Angelstam, 1986). Several studies investigated how edge contrast could mitigate edge effect (Ries et al., 2004), based on the hypothesis that when edge contrast is low, i.e. when the two adjacent habitats have few qualitative or structural differences, then their associated communities are more similar than in high-contrast edges. This similarity can be expressed in terms of total specific diversity and abundance, or within functional groups. Reino et al. (2009) showed a tendency for stronger responses to high-contrast edges (old and tall eucalyptus plantations vs. fallow fields) than to low-contrast edges (young and short oak plantations vs. fallow fields). The same trend was found for forest dung beetles sampled in tropical wood edges, which showed a neutral response for low-contrast edges (mature plantations) but edge avoidance for high-contrast edges (recent plantations) (Peyras, Vespa, Bellocq, & Zurita, 2013). More rarely, similarity between animal or plant communities between forest edges and adjacent open habitat have been estimated with beta diversity indices. Yekwayo, Pryke, Roets, and Samways (2016) showed lower beta diversity and lower species replacement of ground-living arthropods in low contrast edges (natural forests vs. pine plantations) than in high-contrast edges (natural forest vs. grasslands), and Eldegard, Totland, and Moe (2015) showed an increased turnover of species with edge contrast. In addition to local factors, there is a growing number of studies showing that edge effects also depend on the landscape context. Reino et al. (2009) showed that positive edge responses of bird abundances tended to be strongest in less fragmented landscapes, except for steppe birds. In their synthesis, Porensky and Young (2013) conclude that studies from a variety of ecosytems show that edge-effect interactions can have significant consequences for ecosystems and conservation. Moreover a negative effect of management intensity of the surrounding agricultural landscape was demonstrated on
plant diversity in wood edges (Chabrerie, Jamoneau, GalletMoron, & Decoq, 2013). In this study, we investigated the response of plant and pollinator (bees and butterflies) diversity to forest edge contrast. In agricultural landscapes, semi-natural habitats, including forest edges, are an essential source of feeding (flowers and host plants), nesting resources (below-ground and abovegroup bee nesters) for pollinators (Steffan-Dewenter et al., 2006; Morandin & Kremen, 2013) and overwintering sites (Sarthou et al., 2005). In the adjacent habitat, the availability of these resources can depend on habitat persistence. As a consequence, we defined edge contrast level depending on soil disturbance: we considered a forest edge to be low-contrast when it was adjacent to open habitat with permanent vegetation cover and low levels of soil disturbance (i.e. grey dunes, mowed fire-breaks, orchards, grasslands), and high-contrast when the adjacent open habitat showed higher disturbance rates (vegetation removal and ploughing for tilled firebreaks, oilseed rape crops). We tested whether plant and pollinator communities between forest edge and adjacent habitats are more similar in low-contrast edges than in high-contrast edges. We also investigated whether similarity of bee and butterfly communities between forest edge and open habitat depends on edge structure (height and width) and on landscape context: grassland and forest cover, and woody edge length (forest edges plus hedges). Finally, studies investigating edge effects on different biological models that are functionally linked showed cascading effects between various compartments of the ecosystems. Wimp, Murphy, Lewis, and Ries (2011) showed that the decline in specialist herbivores and in their associated predators was driven by an increase in generalist predators near edges. Similarly, Montgomery, Kelly, Robertson, and Ladley (2003) showed that higher fruit set of the mistletoe Peraxilla tetrapetala (Loranthaceae) in edges was due to higher visitation rates of pollinators on edges rather than greater nutrient resource availability. We thus discussed whether the response of plants to edge contrast cascades up to pollinators.
Materials and methods Study sites Forest edges and their adjacent habitats were sampled in three regions of France: Aquitaine (A), Centre (C) and Midi-Pyrénées (MP). These regions have different forest types (intensive forest management/farm forests, conifers/oak species, large/small forests) (Fig. 1 and see http://dynafor.toulouse.inra.fr/data/bilisse/lisieres). The Aquitaine region, on the Atlantic coast in southwestern France, has the largest planted pine forest in Europe – the “Landes de Gascogne” forest – dedicated to wood production. According to the landscape surrounding the studied edges, 67% of the land is covered by maritime pine planta-
E. Andrieu et al. / Basic and Applied Ecology 27 (2018) 83–95
85
Fig. 1. Location of the three regions (Centre, Aquitaine and Midi-Pyrénées). Enlarged views of each region show the location of sampling sites (black dots) in relation to forest cover (gray) and open habitats (white).
tions (Pinus pinaster Aiton) (average size: 5.2 ha), (see van Halder, Barbaro, & Jactel, 2010 for details). The remaining part of the land is covered by herbaceous firebreaks running through. At the Atlantic coast the pine forest is bordered by grey dunes (10% of the studied landscapes). Firebreaks are either mowed or tilled and grey dunes are not managed. The climate is thermo-atlantic (mean annual temperature, 12 ◦ C; mean annual precipitation, 700 mm) and the elevation is low (c. 50 m a.s. l.). The Centre region, south of Paris, is mostly dedicated to intensive crop production (mostly cereals, OSR – oilseed rape – and corn), which covers 47% of the studied landscapes, whereas grassland covers 15%. Forests cover nearly
20% of the the studied landscapes. Most forests are small (average size: 2.2 ha) and are dominated by oak (Quercus petraea and Quercus robur) and hornbeam (Carpinus betulus) coppice-with-standards used for timber and firewood production. The climate is oceanic with slight continental influences (mean annual temperature, 10.6 ◦ C; mean annual precipitation, 640 mm). The Midi-Pyrénées region is situated in the south-west of France. This study was conducted in the Long Term Ecological Research Network Vallées et Coteaux de Gascogne (LTER EU FR 003) (43◦ 17 N, 0◦ 54 E). This hilly region (250–400 m a.s.l.) is characterized by mixed crop-livestock farming system with a mosaic of small forests (26%, average
86
E. Andrieu et al. / Basic and Applied Ecology 27 (2018) 83–95
Table 1. Description of selected forest edges and sample size for each taxonomic group and region (A: Aquitaine, C: Centre, MP: MidiPyrénées). Region
Forest type
Adjacent habitat
Contrast
Anthropic disturbance
Edge sample size Plants
Bees
Butterflies
A
Conifer plantation
Dunes Mowed firebreaks Tilled firebreaks
Low Low High
No disturbance Mowing Tillage
16 4 8
6 6 0
15 8 9
C
Deciduous forests, coppiced
Orchards
Low
11
11
11
Oilseed rape crops
High
Perennial crop with pesticide Conventional farming with tillage, chemical fertilizers and pesticide
10
10
10
Grasslands
Low
15
0
16
Oilseed rape crops
High
8
10
8
MP
Deciduous forests, coppiced
Permanent grasslands with mowing and/or grazing Conventional farming with tillage, chemical fertilizers and pesticide
size: 1.5 ha), grasslands (28%) and crop fields (39%, mainly winter cereals, rapeseed, maize and sunflower) (Choisis et al., 2010). The dominant tree species in the forests are pedunculate oak (Q. robur L.) and sessile oak (Q. petraea Lieblein). The climate is sub-Atlantic with slight Mediterranean influences (mean annual temperature, 12.5 ◦ C; mean annual precipitation, 750 mm).
Site selection We selected 70 sites with a forest edge bordering open habitat (see http://dynafor.toulouse.inra.fr/data/bilisse/lisieres). Within each region, we selected low-contrast edges where forests were adjacent to habitat showing low levels of anthropogenic disturbance (51 in total) and high-contrast edges where forests were adjacent to more intensively disturbed habitats (29 in total). Depending on the region, adjacent habitats of low-contrast edges were grey dunes, mowed firebreaks (region A), orchards (C) or grasslands (MP), and those of high-contrast edges were tilled firebreaks (A) or oilseed rape crops (C and MP). Thirty three sites were selected in region A, 21 in region C and 26 in region MP (Table 1). In each region, forests were selected to be, as far as possible, of the same type (size, tree composition, management). Forest edges were selected to be as straight as possible and at least 100 m long. Only sites with direct contact between forest and open adjacent habitat were considered (no stream, no road or lane). Forest boundaries were defined as the line formed by trees with a diameter of at least 5 cm at breast height (Fig. 2). Edge habitat was defined as the transition zone between wood boundary and the adjacent habitat (Fig. 2). We defined as high-contrast edges, those having adjacent habitats with regular soil disturbance causing a temporary removal of vegetation (tillage in half of the firebreak, tillage and crop harvesting in the rapeseed). The open habitat of low-contrast
Fig. 2. Edge is defined as the transition zone between the wood boundary and the managed adjacent open habitat. It is characterized by the height of the canopy (h1), height of the bottom of tree crown (h2), and the height between the shrub layer and the bottom of the tree crown (h3). Plants (gray squares), butterflies (rectangles) and bees (circles) were sampled in the edges and in the adjacent open habitats.
edges had permanent vegetation cover (no disturbance in dunes, mowing in grasslands, in half of the firebreaks and in orchards) (Table 1).
E. Andrieu et al. / Basic and Applied Ecology 27 (2018) 83–95
87
Table 2. Mean and standard error of explanatory variables used in the multi-inference models to explain plant, bee and butterfly beta diversities. % of grassland, % of woods and Woody linear edge were calculated on a buffer of 500 m radius centered on the butterfly and plant transects. The other variables characterized the edge (h1, h2, h3 and width, see Fig. 2) and the difference in plant communities between the edge and adjacent open habitat (Beta Flora: plant beta diversity for all plant species between the edge and the paired adjacent habitat, Diff Nect: difference in the flower cover between edge and the paired adjacent habitat). Aquitaine
% grass % wood Woody edges (m) h1 (m) h2 (m) h3 (m) Width (m) Diff Nect Beta Flora
Centre
Midi-Pyrénées
All landscapes
Mean
SE
Mean
SE
Mean
SE
Mean
SE
17.9 57.4 4959 12.2 6.0 3.13 20.6 1.20 0.49
2.7 3.2 297 0.53 0.31 0.37 4.4 0.24 0.03
15.1 36.0 6034 18.2 6.2 0.81 7.62 1.24 0.80
2.5 4.2 549 0.70 0.92 0.32 0.61 0.28 0.02
29.9 28.2 8055 17.5 3.2 1.06 5.04 0.33 0.73
2.7 2.5 411 0.90 0.27 0.25 0.54 0.13 0.07
20.6 42.0 6227 15.6 5.3 1.9 12.1 0.94 0.70
1.7 2.4 267 0.50 0.33 0.24 1.99 0.14 0.03
Biodiversity sampling Sampling of bees (Apoidea) and butterflies (Rhopalocera) was conducted in 2011, vascular plant species were recorded in 2011 (region C and partly in region A) and in 2012 (region A, region MP and partly in region A) in the edge and in the adjacent habitats (Table 1, Fig. 2). Plant surveys comprised abundance-dominance records of all vascular plant species according to the Braun-Blanquet scale (Braun-Blanquet 1956). Data from each vegetation layer (herbaceous: 0–0.3 m, shrub: 0.3–7 m and tree: >7 m) were recorded in May–June 2011 or 2012. Surveys were conducted in five 2 × 2 m quadrats regularly arranged along two 100-mlong transects parallel to the border (Fig. 2). The first transect was located at the edge and the second, 20 m into the adjacent open habitat (Alignier, Alard, Chevalier, & Corcket, 2014). Bees were trapped using pan traps (exposed for 15 days in June 2011). Pan traps consisted of two plastic bowls (a yellow and a white one) sprayed with a UV-reflection paint (S.P.R.L. Spray-color 18 133UK, Brussels, Belgium) and mounted on wooden poles at vegetation height in the open habitat (Westphal et al., 2008). Pan traps were filled with a liquid composed of approximately 2.4 L of water, 0.6 L of monopropylene glycol for conservation and a few drops of liquid soap to lower surface tension. To limit inconvenience to the farmers, pan traps were not placed in grazed grasslands and were located at the edge and 10 m away into the other adjacent habitats. Bees were trapped in all regions, except in grasslands in region MP (because of grazing) and in tilled firebreaks in region A (Table 1). All collected bee specimens were stored in a freezer, dried, mounted and identified by the same persons to species level where possible. Butterflies (Lepidoptera: Rhopalocera) were monitored four times a year (in May, June, July and August 2011) along two 100-m-long transects (Pollard & Yates, 1993) placed at the edge and within the adjacent field, 20 m away from the edge. Transect walks (4 min) were undertaken between
10:00 am and 5:00 pm, when weather conditions were suitable for butterfly activity: dry conditions, low wind speed (15 ◦ C) and bright weather. Butterflies were recorded in a fixed width band of 2.5 m on either side of the transect and 5 m ahead. Individuals were determined at the species level, visually or after net-trapping (and release). The total number of individuals per species for the four surveys was scored.
Edge variables On each site, we measured a set of variables to characterise the edge and the adjacent open habitat. Edge structure was described by four variables: h1 (canopy height), h2 (height of the bottom of tree crown), h3 (height between the top of the shrub layer in the understory and the bottom of tree crown) and edge width (length between wood boundary and the border of the adjacent habitat) (Table 2, Fig. 2). The border of the adjacent habitat was defined as the limit of the farming management (for crops: the last seeding line, for grassland: electric fence for grazed grassland or last traces of the mowing for hay meadows). Concomitantly to butterfly surveys, flower cover was visually estimated using the Braun-Blanquet scale along the butterfly transects in the edge and in the adjacent habitat (including rape seed when flowering in region C). We selected the highest cover value of the four visits for each edge and its adjacent habitat, and calculated the difference in nectar resource between them (Diff Nect).
Landscape variables Landscape potential to sustain plants and pollinators was evaluated by quantifying semi-natural elements in a 500 m buffer area, centred on the middle of the edge transect. Landscape variables were: the percentage of buffer area covered by grasslands (%grass) and forests (%wood), and the cumulative length of forest edges and twice the length of hedgerows
88
E. Andrieu et al. / Basic and Applied Ecology 27 (2018) 83–95
(to take into account both sides of the hedges) (Woody edge, in m) (Table 2).
Data analysis The uneven design result from the availability of the suitable sites which varied between region, but also from further practical constraints, among which (i) we were not allowed by owners to leave pan traps in grazed grassland (actually all grasslands in MP), (ii) some firebreaks were mowed and other tilled (region A), (iii) a grassland was converted to wheat between the two years of the study (region MP). This prevent us from testing our hypothesis on all the studied biological models, statistical analyses were done per site and per taxa, with statistical methods robust to uneven sampling size. All statistical analyses were performed with R 3.2.3 (R Development Core Team, 2015). We did not detect any autocorrelation of model residuals (R package ncf, function plot.spine.correlog).
Gamma-like diversity Two gamma-like diversities were calculated for a given type of edge contrast per region: one pooling all the individuals recorded in the edge habitat, and one in the adjacent habitat (e.g. all the butterflies encountered in forest/grassland edges vs. in grasslands of region MP). We compared gammalike diversity between edge and adjacent habitat for each taxon (plants, bees, butterflies) for each type of edge contrast and within each study region using diversity profiles. Diversity profiles provide a faithful graphical representation of the shape of a community; they show how the perceived diversity changes as the emphasis shifts from rare to common species (Leinster & Cobbold, 2012). We used Hill (1973) diversity profiles such as: R q 1/(1−q) q D = pi i=1
where q is the order of diversity and pi = ni /N (ni : number of individuals of species i and N: the total number of individuals). In this study, we used a range of order for diversity from 0 to 2 according to the literature (Marcon, Scotti, Hérault, Rossi, & Lang, 2014): 0 D being species richness (SR), 1 D the exponential of the Shannon Diversity Index (eH ) and 2 D Simpson’s Reciprocal Index (1/D). According to the theory of diversity ordering, one community can be regarded as more diverse than another only if its Rényi diversities are higher all along the curve (no intersection) (Tothmeresz, 1995). The order of diversity for edge and adjacent habitat communities were calculated using the R packages Entropart (Marcon & Héraultt, 2015). A potential sampling bias of existing species was taken into account in the DivProfile function of the Entropart package.
Beta diversity Beta diversities of plants and pollinators between the edge and the paired adjacent habitat were calculated using the Bray–Curtis dissimilarity index upon abundance data. Low values of the Bray–Curtis index indicate that communities have similar species composition and abundance, while high values indicate that they are different. The significance of the difference between beta diversity according to edge contrast was tested using t-test on simple linear modelling between the Bray–Curtis dissimilarity index and the adjacent habitat type. Finally, we tested whether plant beta diversity between wood edge and the adjacent open habitat was related to edge structure (h1, h2, h3, Width) and landscape variables (%grass, %wood, Woody edge). The same analysis was conducted for pollinator beta diversity with an additional variable: the contrast in flora between both habitats. This contrast was estimated using two variables: (i) the plant beta diversity (for all plant species) between the edge and the paired adjacent habitat (Beta Flora) and (ii) the difference in the flower cover between edge and adjacent habitat (Diff Nect). Linear mixed effect models (LMMs) for plant and pollinator beta diversities were fitted with the above variables as fixed factors, and adjacent habitat nested in the region as a random factor. The quality of fit of each LMM was assessed by examining the normality and randomness of the standardized residuals. We then performed a multi-model inference procedure based upon the bias-corrected Akaike’s information criterion (AICc) using the “MuMIn” R package (Barton, 2016). The overall best model and all competing models were ranked in relation to the difference in their AICc scores. Only models with AICc
